A deep Neural Network model was trained to classify the facial expression in unconstrained images, which comprises nine layers, including input layer, convolutional layer, pooling layer, fully connected layers and out...
详细信息
Feature subset selection is an important approach to deal with high-dimensional data. But selecting the best subset of data is NP hard. So most of feature selection methods cannot handle high-dimensional data efficien...
详细信息
Feature subset selection is an important approach to deal with high-dimensional data. But selecting the best subset of data is NP hard. So most of feature selection methods cannot handle high-dimensional data efficiently, or they can only obtain local optimum instead of global optimum. In these cases, when the data consist of both labeled and unlabeled data, semi-supervised feature selection can make full use of data information. In this paper, we introduce a novel semi-supervised feature selection algorithm, which is a filter method based on Fisher-Markov selector, thus ours can achieve global optimum and computational efficiency under certain kernels.
In this paper, we present a new speech enhancement method based on robust principal component analysis. In the proposed method, noisy signal is transformed into time-frequency domain where background noise is assumed ...
详细信息
Electric valve has been applied to various occasions and domains. In some adverse environments where such defects of traditional control system as low efficiency and poor safety have been exposed, optical encoder is a...
详细信息
For natural numbers n and k, where n > 2k, a generalized Petersen graph P(n,k) is obtained by letting its vertex set be {u1,u 2,..,un} ∪ {u1,v2..,u n} and its edge set be the union of vi v i+1,uivi,ViVi+k over 1 ...
详细信息
For natural numbers n and k, where n > 2k, a generalized Petersen graph P(n,k) is obtained by letting its vertex set be {u1,u 2,..,un} ∪ {u1,v2..,u n} and its edge set be the union of vi v i+1,uivi,ViVi+k over 1 ≤ i ≤ n, where subscripts are reduced modulo n. In this paper, an integer programming formulation for Roman domination is established, which is used to give upper bounds for the Roman domination numbers of the generalized Petersen graphs P(n,3) and P(n,4). Together with the dynamic algorithm, we determine the Roman domination number of the generalized Petersen graph P(n, 3) for n ≥ 5.
In view of the draw backs of mango grade identification in China, which still relies on photoelectric sorting and manual separation, this paper presents a processing method on the basis of the technology of computer v...
详细信息
In view of the draw backs of mango grade identification in China, which still relies on photoelectric sorting and manual separation, this paper presents a processing method on the basis of the technology of computer vision and digital image. Utilizing image processing technology, the researcher calculated the length of the long-short-axis, marked the location of it and calculated the 7 parameters, chroma, length, width and etc.,4 of which are chosen as the key characteristics of the BP input of network to build a network and identify the level of mango through analysis of the external characteristics of mango. The method is based on traditional characteristics detection, using boundary tracking algorithm and the length of the new long-short-axis detection algorithm. The result of experiment indicates that the calculating method and judging of the level of mango are precise and accurate, with an average recognition rate of 92%. Therefore, the method has a great practical value, which can be applied to other agricultural products classification.
Point matching is an important component of image registration. Recent years, Coherent Point Drift (CPD) method becomes a very popular point matching approach. CPD treats point matching as a probability estimation pro...
详细信息
Point matching is an important component of image registration. Recent years, Coherent Point Drift (CPD) method becomes a very popular point matching approach. CPD treats point matching as a probability estimation problem and speeds up the process of matching a lot. In this method, one set of points are thought to be sampled from a Gaussian Mixture Model (GMM), which is centered by the other set of points. However, CPD is sensitive to outliers and noises, especially when the noise ratio increased or the number of outliers gets much high. To deal with this problem, we introduce shape context into the step of searching for matching points and then improve the form of prior probabilities of GMM in this paper. The main idea of our method is that if the most points in a data set are likely to be matched to a particular centroid, this Gaussian component should be have a more influence to GMM. Therefore, we set prior probability of GMM with the similarity between GMM components and the data set. And the computation of similarity is based on shape context. The experiments on 2D and 3D images show that when noise ratio is low, our method performs as well as CPD does, but as the ratio increased, our method is more robust and satisfactory than CPD.
Optimization in noisy environments is regard as a favorite application domains of genetic algorithms. Different methods for reducing the influence of noise are presented and discussed. A new fitness evaluation method ...
详细信息
ISBN:
(纸本)9781479969302
Optimization in noisy environments is regard as a favorite application domains of genetic algorithms. Different methods for reducing the influence of noise are presented and discussed. A new fitness evaluation method is proposed that reevaluates all survival individuals each generation. Compared with re-sampling and population sizing, the new evaluation approach shows higher probability of searching to the global extremum area and precision of convergence. These results demonstrate that the proposed method is effective for reducing noise effects.
The existing safety and health monitoring methods for bridge construction are mainly manual monitoring and wired monitoring with many disadvantages, such as low efficiency, poor accuracy, great implementation difficul...
详细信息
For multi-target route optimization with constraint conditions, the mathematical model for logistics distribution route optimization is built to accelerate response speed of logistics enterprises to customers, improve...
详细信息
For multi-target route optimization with constraint conditions, the mathematical model for logistics distribution route optimization is built to accelerate response speed of logistics enterprises to customers, improve service quality, and strengthen the satisfaction of customers, and a new algorithm with the combination of genetic and ant colony algorithms is proposed to solve the selection issues of such logistics route. Initial pheromone is formed with genetic algorithm, based on which the optimal solution is rapidly sought with ant colony algorithm, and complementary advantages are achieved between above two algorithms. Application examples and simulations are available for calculation, and the results show that such algorithm is practical and effective to optimize logistics distribution route.
暂无评论